Prepare Generator Output for Evaluation

Transforms the output from generate_weather into the format required by evaluate_weather_generator. This helper function handles the date resampling, complete-year filtering, and data extraction that is typically needed between generation and evaluation steps.

Usage

prepare_evaluation_data(
  gen_output,
  obs_data,
  obs_dates,
  grid_ids,
  variables,
  min_days_per_year = 365L,
  verbose = TRUE
)

Arguments

gen_output

List returned by generate_weather, containing resampled (tibble of resampled observation dates per realization) and dates (vector of simulated dates).

obs_data

Named list of data frames from read_netcdf$data, one per grid cell. Each data frame must contain columns for all variables.

obs_dates

Date vector corresponding to rows in each obs_data element. Typically from read_netcdf()$date.

grid_ids

Character or integer vector of grid cell identifiers to include in the evaluation. Must match names or indices in obs_data.

variables

Character vector of variable names to extract (e.g., c("precip", "temp")).

min_days_per_year

Integer. Minimum number of days required to consider a year complete. Default is 365. Use 360 for 360-day calendars.

verbose

Logical. If TRUE, prints progress messages to console via the internal .log() function. Default is TRUE.

Value

A list with two elements:

sim_data

List of length n_realizations. Each element is a list of data frames (one per grid cell) with columns date followed by the requested variables.

obs_data

List of data frames (one per grid cell) with columns date followed by the requested variables.

Details

The function performs the following transformations:

1. Maps resampled dates back to row indices in the original observations

2. Filters the simulation period to complete years only (>= 365 days)

3. Extracts the specified variables from observed data using resampled indices

4. Formats both simulated and observed data as required by the evaluator

Complete years are identified by calendar year boundaries. For water-year simulations, users should ensure the simulation spans full water years.

See also

generate_weather, evaluate_weather_generator

Examples

if (FALSE) { # \dontrun{
# After running generate_weather()
ncdata <- read_netcdf("climate_data.nc")
gen_output <- generate_weather(
  obs_data = ncdata$data,
  obs_grid = ncdata$grid,
  obs_dates = ncdata$date,
  ...
)

# Prepare data for evaluation
eval_data <- prepare_evaluation_data(
  gen_output = gen_output,
  obs_data   = ncdata$data,
  obs_dates  = ncdata$date,
  grid_ids   = ncdata$grid$id,
  variables  = c("precip", "temp")
)

# Run evaluation
results <- evaluate_weather_generator(
  daily_sim = eval_data$sim_data,
  daily_obs = eval_data$obs_data,
  ...
)
} # }